Crystalline Silicon Solar Cells with Segmented Selective Emitter byUltraviolet Laser Doping

John S. Renshaw∗†, Ajay Upadhyaya∗, Vijaykumar Upadhyaya∗ and Ajeet Rohatgi∗‡

∗University Center of Excellence for Photovoltaics, Georgia Institute of Technology, Atlanta, GA, 30332, U.S.A.†Department of Physics, Georgia Institute of Technology, Atlanta, GA, 30332, U.S.A.

‡Suniva, Norcross, GA,30092 , U.S.A.

Abstract—A solar cell design with a UV laser doped segmentedselective emitter is reported. Several different laser settings areexplored to determine the optimum power for this process andit is found that the pitch between the segmented n++ regions iscritical to the short circuit current (Jsc) of the cell. An increaseof 0.4 mA/cm2 is seen in the Jsc when the pitch is increased from50 µm to 200 µm while maintaining the fill factor of 79%.

I. INTRODUCTION

Laser doping for selective emitter formation in screenprinted crystalline Silicon solar cells is an elegant method toachieve low contact resistivity and high emitter sheet resistancesimultaneously, which can increase cell efficiency by up to0.5 % absolute [1]. The selective emitter concept is attractivebecause it has low doping levels between the contacts givingthe device a low reverse saturation current density in theemitter resulting in high Voc relative to non selective emittercells as well as higher doping levels beneath the contacts pro-viding low contact resistance and high FF. Selective emitterscan be made in a variety of ways, however some of thesemethods require multiple steps and are not economical forindustrial use. There are several selective emitter approachesthat are under industrial development for screen printed solarcells including doped silicon ink [2], implantation [3], emitteretch back [4] and laser doping [5].The laser doping methodis unique in that it utilizes phosphosilicate glass (PSG) fromthe POCl3 diffusion as a dopant source for selective emittercreation beneath the Ag contacts. To the best of our knowledgethe selectively doped region is not segmented in other works,it either runs the full length of the finger or is perpendicularto the fingers running the length of the wafer to achieve easieralignment [6]. In this work we report on an alternative methodfor creating the laser doped selective emitter (LDSE) solar cellusing an ultraviolet laser and a segmented selective emitterdesign. This design has the potential of giving higher Jsc andopen circuit voltage (Voc) because of the lower volume ofhigh Auger recombination and recombination through laserinduced defects compared with a non segmented selectiveemitter design.

II. EXPERIMENT

The laser employed for this research is a pulsed, frequencytripled Coherent Avia laser with a wavelength of 355 nm. Ascan head is used to raster the beam across the wafer surfacein the desired pattern, the scan speed across the wafer is 3000

mm/s and the laser spot has an 50 µm diameter and is gaussianin shape. In this work we used commercial grade 2 Ω-cm,239 cm2, Czochralski (CZ) grown Si, P type wafers. Thefabrication procedure for LDSE cells is shown in figure 1.This process only varies slightly from a standard industrial

Fig. 1. Process flow for LDSE solar cells. Step 1: Damage Etch andtexture wafer surface Step 2:POCl3 diffusion to create the n+ emitter Step3:Laser doping in selective regions to create the n++ regions Step 4:Finishthe solar cell by chemical edge isolation, PSG removal, PECVD SiN for theantireflection coating,screen printing and firing contacts

solar type cell process flow, the only additional step is laserdoping. The first step in the LDSE process is to remove thesaw damage and texture the surface with pyramids to reducereflectance. This is followed by diffusion in a tube furnaceusing POCl3 as a P source. After diffusion the Si surfaceis now n+ with a PSG coating. After diffusion laser dopingbeneath the grid lines drives in P dopants from the PSG glassin 500 µm long strips that are spaced about 100 µm center tocenter and have the width of the laser spot size. The resultingdoping profiles from POCl3 diffusion and laser doping forvarious laser conditions can be seen in figure 2. Followinglaser doping the wafers go through chemical edge isolation and

978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001098978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001098

PSG removal in HF. A commercial PECVD SiN tool is thenused to form the antireflection coating on the wafers beforeAg and Al paste are screen printed on the front and rear of thecell respectively. Finally the wafers are fired in a belt furnaceto form the contacts and create the p+ back surface field. TableI summarizes the IV parameters of the best and average cellsfrom a run of 12 wafers.

TABLE IAVERAGE AND BEST IV CHARACTERISTICS, MEASURED UNDER

STANDARD TEST CONDITIONS

Voc(mV) Jsc(mA/cm2) F.F. (%) Efficiency (%)Average 632 36.9 79.1 18.4Best 633 36.9 79.4 18.6

Fig. 2. ECV Doping profiles for: the POCl3 emitter before laser doping,laser doping profiles for various laser powers using PSG a dopant source andthe result of laser irradiation with no P source

III. RESULTS AND DISCUSSION

As previously mentioned this process differs from otherspresented in literature because the laser used in this researchemploys a UV laser where as most others have used a greenlaser and this process uses a segmented selective emitter. Thedoping profiles generated from the UV laser seen in figure 2are quite different from those obtained by laser doping with agreen laser [1]. The deepest junction formed is only 0.4 µmdeep where as junction depths of 0.8-1 µm can be formedusing a green laser. This result is not surprising since theabsorption coefficient for 355 nm uv light in Si is over twoorders of magnitude larger than for 532 nm green light [7].The shallower junction depth provided by UV laser dopingcould be advantageous because the shallower volume of highdoping where Auger recombination is high. The total activeP dose for each profile was calculated by integration and isshown in table II

TABLE IIACTIVE P DOSE FOR EACH PROFILE. OBTAINED BY INTEGRATING THE

ECV DATA.

Profile Dose (1015 atoms/cm2)90 Ω/ POCl3 emitter 1.54180 khz, 33 µJ/pulse laser doping 2.13180 khz, 33 µJ/pulse laser doping(PSG removed)

1.09

150 khz, 50 µJ/pulse laser doping 2.52120 khz, 86 µJ/pulse laser doping 3.23

It is interesting to note that the total P dose increasesbetween 38% and 109% of the starting POCl3 dose dependingon the laser power used but when the PSG is removed beforelaser doping, the total P dose goes down. This shows that thePSG is the source of the extra P dose incorporated into theprofile and not inactive dopants near the surface. The fact thatthe dose goes down after laser doping without the PSG couldbe explained if some of the P atoms end up in interstitial sitesafter laser doping. The peak doping is higher in samples dopedwith the 180 khz rep. rate compared with samples doped withthe 150 khz and 120 khz rep. rate, this is because the powerassociated with the 180 khz rep. rate is insufficient to melt theSi, where as at 150 khz the pyramids start to melt and at 120khz they are completely gone and the surface is flat. Once theSi is melted P can diffuse quickly and redistribute the highdoping level at the surface due to the high diffusivity of P inSi melt [8]. For this reason we chose the 180 khz rep. rate inour process.

Even though the surface concentration for the laser dopedselective emitter is lower than the POCl3 emitter we see alower contact resistance on the laser doped samples indicatedby the average cell series resistance that is three times loweron samples that are laser doped compared with reference cellsmade on the same emitter with no laser doping ( 0.6 vs 2.0 Ω-cm2). This is because screen printed Ag contacts make contactto Si through Ag crystallites that etch into Si on average0.13 µm [9]. At this depth the POCl3 emitter doping levelis approximately 3.8x1018/cm3 where as the 180 khz rep. ratedoped profile has a P concentration of about 4x1019/cm3. Thisconcept is demonstrated in figure 3, which shows the micro-scopic contact resistivity between the silver crystallites andthe substrate for these different doping profiles as a functionof depth. The microscopic contact resistivity was obtainedusing the Wentzel-Kramers-Brillouin [10] approximation inthe thermionic field emission regime, a Schottky barrier heightof 0.78 eV was used in the calculations. Where the dashedline drawn at ρc=10−3mΩ -cm2 intersects each doping profilerepresents the maximum depth at which good contact betweenthe Ag crystallite and the emitter can be formed [1]. For thePOCl3 emitter profile most of the Ag crystallite would bemaking contact to lowly doped Si, however for the laser dopedsamples the entire Ag crystallite would make good contact tothe Si, explaining the better contact in laser doped samples. Itshould be noted that this contact resistance does not represent

978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001099978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001099

Fig. 4. LBIC map showing the response at 980 nm for a representative finger on solar cells with 50 and 100 µm selective emitter pitch

the actual contact resistance of the whole Ag finger to theemitter but rather an approximation of the contact resistance ofthe Ag crystallites to the emitter formed during contact firing.

Fig. 3. Microscopic contact resistivity using the Wentzel-Kramers-Brillouinapproximation for the thermionic field emission regime

The effect of the pitch of the selective emitter is demonstratedin table III. The benefit of the segmented selective emitteris seen in the short circuit current (Jsc) which increases asthe pitch between the selective emitter regions increases. Thiseffect does not appear to be due to a change in the reflectanceof the wafers, the laser power used in the experiment does notalter the pyramidal texturing of the wafers. The authors believethe current is most likely lower in samples with smaller pitchto due to traps created from laser induced defects, this effectis currently under investigation.

The lower response region around the contacts due to Augerrecombination and laser induced defects can be seen in laserbeam induced current (LBIC) maps, shown in figure 4. Forsamples with a 50µm selective emitter pitch, a fairly wide

TABLE IIIAVERAGE AND IV CHARACTERISTICS FOR VARYING SELECTIVE EMITTERPITCH, MEASURED UNDER STANDARD TEST CONDITIONS. THE SELECTIVEEMITTER REGION IS 500µM WIDE AND ABOUT 50µM THICK, THE LONGERSIDE IS THE ONE PERPENDICULAR TO THE FINGER, 180 KHZ, 33 µJ/PULSE

LASER SETTING WAS USED ON THESE WAFERS

Pitch Voc(mV) Jsc(mA/cm2) F.F. (%) Efficiency (%)No SelectiveEmitter

629 37.3 72.4 17.0

50µm (over-lapping)

629 36.7 79.5 18.4

100µm 629 37.0 79.7 18.5200µm 629 37.1 79.0 18.5

region of lower response can be seen around the finger. As thepitch increases to 100 the area of the lower response regiondecreases. It can be seen from the correscan plots in figure5 that the line contact resistance of samples with 50 to 100µm pitch are very low and have roughly the same contactresistance. This is impressive since the selective emitter cov-erage is roughly half for the 100 µm pitch compared to the 50µm pitch. For samples with a 200 µm pitch selective emitter,line contact resistance is still good; however, there are a fewlocalized regions with higher line contact resistance. Theseregions are most likely the cause of the slightly lower fillfactor for samples with a 200 µm pitch selective emitter seenin table III. Without the selective emitter the contact resis-tance is highly non-uniform and is not sufficient for high fillfactors. Segmented selective emitters allow for improvementin response around the grid line compared to a homogeneousselective emitter without any sacrifice in the contact resistance.

IV. CONCLUSION

In conclusion, we have demonstrated that an ultraviolet lasercan be used to create a selective emitter in the fabrication ofhigh efficiency crystalline Si solar cells. The selective emitter

978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001100978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001100

Fig. 5. correscan of cells with no selective emitter and with selective emitterpitch of 50,100 and 200µm.

profiles created are suitable for low-ohmic contacts and thatthe dopant source is the PSG not the inactive dopants in theelectrically dead layer at the surface. We find that deviceperformance is best when the selective emitter is segmentedand not continuous due to an increase in the Jsc. Segmentedselective emitters have low contact resistance even though onlya fraction of the finger is contacting the selective emitter.We also show that segmented selective emitter solar cells canhave a more uniform response than a non-segmented selectiveemitter solar cell. This work is supported by the United StatesDepartment of Energy.

REFERENCES

[1] T. Roder, S. Eisele, P. Grabitz, C. Wagner, G. Kulushich, J. Kohler, andJ. Werner, “Add-on laser tailored selective emitter solar cells,” Progressin Photovoltaics: Research and Applications, vol. 18, no. 7, pp. 505–510, 2010.

[2] H. Antoniadis, F. Jiang, W. Shan, and Y. Liu, “All screen printedmass produced silicon ink selective emitter solar cells,” in PhotovoltaicSpecialists Conference (PVSC), 2010 35th IEEE, june 2010, pp. 001 193–001 196.

[3] R. Low, A. Gupta, N. Bateman, D. Ramappa, P. Sullivan, W. Skinner,J. Mullin, S. Peters, and H. Weiss-Wallrath, “High efficiency selectiveemitter enabled through patterned ion implantation,” in PhotovoltaicSpecialists Conference (PVSC), 2010 35th IEEE. IEEE, 2010, pp.001 440–001 445.

[4] A. Dastgheib-Shirazi, H. Haverkamp, B. Raabe, F. Book, and G. Hahn,“Selective emitter for industrial solar cell production: a wet chemicalapproach using a single side diffusion process,” in 23rd EuropeanPhotovoltaic Solar Energy Conference, 2008.

[5] Ainhoa, F. Esturo-Bretn, M. Binaie, M. Breselge, T. Citro, X. Frie,M. Gay, D. Geiger, S. Hammer, J. Hsu, J. Hung, T. Isenberg, J. Kuhn,S. Olkowska-Otzel, J. Keller, P. Haase, and Fath, “Selective emitterby laser diffusion on c-si solar cells in industrial high efficiency massproduction,” in Photovoltaic Specialists Conference (PVSC), 2010 37thIEEE. IEEE, 2011.

[6] L. Zhu, J. Gong, J. Huang, P. She, M. Zeng, L. Li, M. Dai, andQ. Wan, “Improving the efficiency of crystalline silicon solar cells by

an intersected selective laser doping,” Solar Energy Materials and SolarCells, 2011.

[7] M. Green and M. Keevers, “Optical properties of intrinsic silicon at 300k,” Progress in Photovoltaics: Research and Applications, vol. 3, no. 3,pp. 189–192, 1995.

[8] H. Kodera, “Diffusion coefficients of impurities in silicon melt,”Japanese journal of applied physics, vol. 2, no. 4, pp. 212–219, 1963.

[9] C. Ballif, D. M. Huljic, G. Willeke, and A. Hessler-Wyser, “Silver thick-film contacts on highly doped n-type silicon emitters: Structural andelectronic properties of the interface,” Applied Physics Letters, vol. 82,no. 12, pp. 1878 –1880, mar 2003.

[10] A. Yu, “Electron tunneling and contact resistance of metal-siliconcontact barriers,” Solid-State Electronics, vol. 13, no. 2, pp. 239–247,1970.

978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001101978-1-4673-0066-7/12/$26.00 ©2011 IEEE 001101